How It Works
Probabilistic model
A model that answers with likelihoods rather than single certain answers — “80% chance this email is spam” instead of simply “spam.” Most machine learning is probabilistic underneath, including the way language models pick each next word. Bayes’ theorem is the classic rule for updating those likelihoods as new evidence arrives.
Origin · no single documented coiner
Grows out of probability theory and statistics going back to Thomas Bayes and Pierre-Simon Laplace in the 18th century; no single coiner of the phrase.